Lexicon
Gauge
A gauge is a metric that snapshots a current value that rises and falls: memory in use, queue depth, connections. How to aggregate gauges without lying.
Definition
A gauge is a metric type that records a point in time value that can move in both directions: memory in use, queue depth, active connections, temperature, items in a cart. Unlike a counter, the current reading is the meaningful number.
What it means in observability
Gauges describe state rather than events. They answer how full, how many right now, how hot, which makes them the language of saturation and capacity: the disk approaching full, the connection pool nearing its limit, the queue backing up. Much of infrastructure monitoring is gauges, and the interesting analysis is usually about levels and trends, is this climbing toward a ceiling, rather than the rates that dominate counter analysis.
How it works in practice
A gauge series is a sequence of snapshots, and aggregation happens across two dimensions. Across time, min, max, average, or last summarize a window, with last available for reading the current value without averaging away recency. Across series, aggregating over hosts or pods needs intent: the average memory across a fleet, the maximum queue depth on any instance, and the sum of connections overall are three different questions with three different answers.
Where it gets hard
Rate style functions applied to gauges produce meaningless jitter, since gauges do not accumulate. Averaging across instances is the subtler trap: fleet average memory at 60 percent hides the one node at 98, so saturation questions usually want max or high percentiles. Sparse reporting adds gaps that interpolation can paper over misleadingly, and a gauge that stops reporting can look healthy precisely when its emitter has died.
Where Tsuga fits
Tsuga's explorer identifies gauges and their sensible aggregations, including a last function for current value reads, so fleet questions get asked with the aggregation they actually mean.
Related terms
- CounterA counter is a metric type for values that only accumulate: requests served, errors thrown, bytes transmitted, jobs completed.
- HistogramA histogram is a metric type that captures how values distribute, by counting observations into buckets, rather than recording what any single value was.
- Infrastructure monitoringInfrastructure monitoring is the practice of tracking the health, performance, and capacity of the layer applications run on: hosts, virtual machines, containers, Kubernetes clusters, and managed cloud services.
- MetricA metric is a named numeric measurement tracked over time, request counts, memory usage, response latency, carrying a type, a unit, and attributes that slice it into series.